Observability is one of those things nobody budgeted for until it suddenly became the line item that ate your infrastructure costs. Datadog's pricing model — $15 per host for infrastructure monitoring, $30+ for APM, extra for logs, traces, synthetic checks, and custom metrics — means a 10-server fleet with full-stack monitoring easily costs $500-$1,500 per month. For startups and small teams, that's rent. The good news is that the open-source observability stack has matured to the point where you can build a Datadog-equivalent setup yourself for the cost of a single $10 VPS to run it on.
The core stack is what the CNCF calls the 'PLG stack': Prometheus for metrics, Loki for logs, and Grafana for visualization, with Tempo or Jaeger bolted on for distributed tracing. Prometheus scrapes metrics from your applications via a simple HTTP /metrics endpoint, stores them in a time-series database, and exposes them for querying via PromQL. Loki does the same for logs — it indexes only metadata (labels) and stores the log lines themselves in cheap object storage, which makes it dramatically cheaper than Elasticsearch for the same volume. Grafana ties them together with dashboards that can combine metrics, logs, and traces in a single pane.
Setting it up is more approachable than it sounds. A docker-compose.yml with five services (Prometheus, Loki, Grafana, Tempo, and a node-exporter for host metrics) gets you a working observability stack in under an hour. Grafana Labs publishes pre-built dashboards for everything you'd ever want to monitor: Node.js apps, Postgres, Redis, Nginx, Kubernetes, Docker — all available as one-click imports from grafana.com/dashboards. For alerting, Grafana's built-in alert manager handles threshold-based alerts and routes them to Slack, PagerDuty, email, or any webhook. The result is a monitoring setup that's functionally indistinguishable from Datadog for 95% of use cases.
Where Datadog still wins is the setup-and-forget factor. Their agents auto-discover services, instrument popular libraries out of the box, and the APM tracing works with zero configuration for most languages. With the open-source stack, you'll spend a weekend writing dashboards and configuring exporters. But that upfront investment pays off: once it's running, you own it forever, there are no per-host fees, and you can monitor 100 servers as easily as 1. For teams that have already been billed $500+ by Datadog for a month of monitoring, the math is brutal — that's a year of VPS costs for the self-hosted equivalent.
A few practical tips if you go this route. First, use Grafana Cloud's free tier as a stepping stone — it gives you 10,000 series of Prometheus metrics, 50GB of logs, and 50GB of traces for free, which is enough for a small project. When you outgrow it, you can self-host the same stack with the same dashboards. Second, for application tracing, use OpenTelemetry — it's the open standard that Datadog, New Relic, Honeycomb, and Grafana Tempo all support, so you can switch backends without re-instrumenting your code. Third, if log volume is your problem, pair Loki with Promtail's pipelined filtering to drop noisy logs before they hit storage — a 10x reduction is typical for most apps.
The observability market is in the middle of a reformation. Datadog's revenue keeps climbing, but so does adoption of open-source alternatives. The PLG stack isn't quite a drop-in Datadog replacement — you'll spend more time on setup and dashboards — but for any team that values data ownership, predictable costs, and the ability to monitor without a credit card on file, it's the right answer. Start with Grafana Cloud's free tier, migrate to self-hosted when you hit the limits, and watch your monitoring bill go from $500/month to $10/month in the process.